Early Screening of Autism Spectrum Disorder ... · mote earlier diagnosis. The current review indicates that screening in children aged 18 to 24 months can assist in early detection,

Early Screening of Autism Spectrum Disorder:Recommendations for Practice and Research

abstractThis article reviews current evidence for autism spectrum disorder(ASD) screening based on peer-reviewed articles published to Decem-ber 2013. Screening provides a standardized process to ensure thatchildren are systematically monitored for early signs of ASD to pro-mote earlier diagnosis. The current review indicates that screeningin children aged 18 to 24 months can assist in early detection, consis-tent with current American Academy of Pediatrics’ recommendations.We identify ASD-specific and broadband screening tools that havebeen ev-aluated in large community samples which show particularpromise in terms of accurate classification and clinical utility. We alsosuggest strategies to help overcome challenges to implementing ASDscreening in community practice, as well as priorities for future re-search. Pediatrics 2015;136:S41–S59

AUTHORS: Lonnie Zwaigenbaum, MD,a Margaret L.Bauman, MD,b Deborah Fein, PhD,c Karen Pierce, PhD,d

Timothy Buie, MD,e Patricia A. Davis, MD,f CraigNewschaffer, PhD,g Diana L. Robins, PhD,g Amy Wetherby,PhD,h Roula Choueiri, MD,i Connie Kasari, PhD,j Wendy L.Stone, PhD,k Nurit Yirmiya, PhD,l Annette Estes, PhD,m

Robin L. Hansen, MD,n James C. McPartland, PhD,o MarvinR. Natowicz, MD, PhD,p Alice Carter, PhD,q DoreenGranpeesheh, PhD, BCBA-D,r Zoe Mailloux, OTD, OTR/L,FAOTA,s Susanne Smith Roley, OTD, OTR/L, FAOTA,t andSheldon Wagner, PhDu

aDepartment of Pediatrics, University of Alberta, Edmonton,Alberta, Canada; bDepartment of Anatomy and Neurobiology,Boston University School of Medicine, Boston, Massachusetts;cDepartment of Psychology, University of Connecticut, Storrs,Connecticut; dDepartment of Neurosciences, University ofCalifornia San Diego, La Jolla, California; eHarvard MedicalSchool and Massachusetts General Hospital for Children, Boston,Massachusetts; fIntegrated Center for Child Development,Newton, Massachusetts; gA.J. Drexel Autism Institute, DrexelUniversity, Philadelphia, Pennsylvania; hDepartment of ClinicalSciences, Florida State University College of Medicine,Tallahassee, Florida; iDivision of Developmental and BehavioralPediatrics, University of Massachusetts Memorial Children’sMedical Center, Worcester, Massachusetts; jGraduate School ofEducation & Information Studies, University of California LosAngeles, Los Angeles, California; kDepartments of Psychology andmSpeech and Hearing Sciences, University of Washington, Seattle,Washington; lDepartment of Psychology, Hebrew University ofJerusalem Mount Scopus, Jerusalem, Israel; nDepartment ofPediatrics, University of California Davis MIND Institute,Sacramento, California; oYale Child Study Center, New Haven,Connecticut; pGenomic Medicine Institute, Cleveland Clinic,Cleveland, Ohio; qDepartment of Psychology, University ofMassachusetts, Boston, Massachusetts; rCenter for Autism andRelated Disorders, Tarzana, California; sDepartment ofOccupational Therapy, Thomas Jefferson University, Philadelphia,Pennsylvania; tUSC Mrs T.H. Chan Division of Occupational Scienceand Occupational Therapy, Los Angeles, California; anduBehavioral Development & Educational Services, New Bedford,Massachusetts

(Continued on last page)

PEDIATRICS Volume 136, Supplement 1, October 2015 S41

SUPPLEMENT ARTICLE

by guest on May 8, 2020www.aappublications.org/newsDownloaded from

Although there have been consider-able advances in characterizing earlybehavioral markers predictive of au-tism spectrum disorders (ASDs), assummarized in this special issue toPediatrics,1 translation into clinicalpractice requires that the process ofmonitoring for such early risk mark-ers be operationalized to facilitatebroad implementation. To that end,universal screening for ASD has beenrecommended by the AmericanAcademy of Pediatrics (AAP) to ensureconsistent practice and optimal de-tection of young children with earlysigns of ASD across a range of clinicaland community contexts.2 The AAP hasrecommended that all children bescreened with an ASD-specific in-strument during well-child visits atages 18 and 24 months in conjunctionwith ongoing developmental surveil-lance and broadband developmentalscreening. The rationale for this rec-ommendation was based on thepresence of ASD symptoms by age 18months, promising data on early ASD-screening tools, and the availability ofeffective intervention strategies tar-geting this age group.3,4 Recent ran-domized controlled trials have addednew evidence that for many childrenaged,3 years, early intervention canimprove outcomes, including coredeficits of ASD (ie, social attention), IQ,language, and symptom severity,5,6

thus increasing the potential benefitsof early diagnosis facilitated by earlyscreening.

Some scientists and practitioners havequestioned whether the evidence rela-tive to general developmental surveil-lance warrants ASD screening,7,8 andothers have argued that research needsto move beyond risk classification andevaluate longer term outcomes of ASDscreening (eg, impact on age of di-agnosis, related gains attributable toearlier enrollment in intervention).9 Theuptake of ASD screening into pediatric

practice has been modest.10,11 Althoughpotential facilitatorsand barriers toASD screening have been researchedanddebated,11–13 screening rates inmanyregions of the United States remain low.Community-based interventions aimedat implementing or increasing utiliza-tion of ASD screening have emphasizedtraining primary care physicians andtheir front-line staff, providing ongoingtechnical assistance (eg, scoring, datamanagement support), and clear re-ferral pathways for specialized assess-ments.9,11,14–17 However, ongoing debateregarding whether there is sufficientevidence in support of ASD screening towarrant widespread practice change8,18

may undermine the degree to whichcommunity pediatricians are adoptingthe AAP policy.

Thus, an updated literature review andbest practice recommendations re-garding ASD screening are warranted,aswell as furtherconsiderationsofhowto address potential barriers to uptakeof screening into clinical practice. Tothat end, an international multidisci-plinary panel of clinical practitionersand researchers with expertise in ASDand developmental disabilities wasconvened in Marina del Rey, Californiain October 2010. The panel reachedconsensus on “How can we optimizedevelopmental course and outcomesthrough ASD screening programs forchildren aged #24 months?”

For further context, we briefly defineterms used to describe the classifica-tion accuracy of specific screeningmeasures. “Sensitivity” refers to theproportion of children with ASD whoare correctly identified as “high risk”according to results of screening;a child with ASD who is not identified bythe screen is considered to be a false-negative. Specificity refers to the pro-portion of children who do not haveASD who are correctly classified usingthe screening tool as not having riskfor ASD; a child who does not have ASD

yet screens positive is considered to bea false-positive. It has been suggestedthat to even receive consideration forpopulation screening applications, thesensitivity and specificity of a screen-ing tool should exceed 0.70.19 However,the relative “cost” associated withfalse-positive and false-negative find-ings, as well as the prevalence of thecondition being screened, must also betaken into consideration. The positivepredictive value (PPV) for ASD ofa screening test is defined as the pro-portion of children screening positivewho receive an ASD diagnosis dividedby the total number of screen-positivecases. The negative predictive value(NPV) is the proportion of screen-negative children not receiving anASD diagnosis. PPV and NPV are influ-enced by the baseline prevalence ofASD in the population being screenedas well as the sensitivity and specificityof the screening tool. Although sensi-tivity and specificity are intrinsicmeasures of test performance, PPVand NPV arguably have more inherentmeaning for individual family-level andsystem-level evaluations of screening.

It is also important to distinguish level 1from level 2 screening. Level 1 screen-ing applies to all children regardless ofrisk status (ie, “universal” screening).In contrast, level 2 screening is tar-geted at children already identified asbeing at increased risk (eg, due toa positive family history, concernsraised by parents or clinicians, identi-fication by a level 1 screener).

METHODS

The working group co-chairs and panelco-chairs conducted a PubMed searchto identify relevant articles on screen-ing for ASD in children aged #24months. Members of the workinggroup reviewed the articles. Weassessed whether tools were beingevaluated in the population in whichthey were being considered for use and

S42 ZWAIGENBAUM et al by guest on May 8, 2020www.aappublications.org/newsDownloaded from

whether these tools met the minimumcriteria for specificity and sensitivity tosupport implementation in the generalcommunity. Panel recommendationswere based on this evaluative frame-work.

The working group summarized pub-lished research on screening tools de-veloped for use in children aged #24months, even if the age range of thesescreens exceeded 2 years (Table 1). APubMed search was conducted on June30, 2010, by using the search terms(“child developmental disorders, perva-sive” or “autistic disorder/” or “autism[tw]” or “autistic [tw]”) and (“massscreening” or “screen [tw]”), with theage filter (“infant, birth-23 months”) andlimited to English-language articles. Thissearch yielded 111 references, whichwere reviewed by Drs Zwaigenbaum andBauman, who selected articles focusingon studies that involved prediagnosticscreening for early behavioral orbiological features (as opposed topostdiagnostic screening for etiologicfactors or associated comorbidities).The search results were complementedby additional publications identified byworking group members. Thus, al-though the search strategy was com-prehensive, selection of articleswas notsystematic, which is an important limi-tation. A scoping approach was usedinstead, with some discretion of themultidisciplinary expert working group,to select articles of highest relevance.

Most of the instruments reviewed weredesigned to identify children at risk forASD who warranted further evaluation.Also reviewed were general develop-mental, or broadband, screeninginstruments that had been evaluatedfor the purpose of early identification ofASD, even if not specifically designed todistinguish risk for ASD from risk forother developmental delays. We alsodistinguished between the instrumentsthat had been evaluated as level 1screens, level 2 screens, or both.

During the conference, the workinggroup offered draft recommendationsfor discussion, modification, and rati-fication by all attendees. Electronicvoting was used to express opinionsand guide consensus building. Amodifiednominal group techniquewasused to review the recommendations,with consensus reached by$1 roundof voting. The consensus statementsand discussion were summarized asdraft proceedings of the conference,which were subsequently edited by allparticipants. The search was updatedby using the same strategy to addarticles published to December 31,2013, which yielded an additional 85references; selection was limited toprediagnostic screening of early be-havioral or biological markers. Theworking group reviewed and ap-proved the final wording of the sum-mary and recommendations.

The measurement properties thatcharacterize the accuracy of screen-ing instruments used to identifychildren at risk for ASDs are sum-marized in Table 1.17,20–47 ASDscreeners with published evaluationdata include parent questionnairessuch as the Modified Checklist forAutism in Toddlers (M-CHAT),24 theQuantitative Checklist for Autism inToddlers (Q-CHAT),33 the Early Screen-ing of Autistic Traits questionnaire(ESAT),22,23 and the First Year Inventory(FYI).20,48 Table 1 also summarizes ASDscreening instruments with only pre-liminary data (eg, the Pervasive De-velopmental Disorders Rating Scale),36

which will not be included in thepresent discussion.

The results of the overall process arelisted as summary statements. Some ofthe statements summarize the state ofthe literature, whereas others providerecommendations for research neededto fill important evidence gaps and/oraddress issues important for clinicalpractice.

SUMMARY STATEMENTS

Statement 1: Evidence supports theusefulness of ASD-specificscreening at 18 and 24 months.ASD screening before 24 monthsmay be associated with higherfalse-positive rates than screeningat ‡24 months but may still beinformative.

ASD-specific screening in childrenaged 18 to 24 months can assist inearly detection

Table 1 summarizes the measurementproperties of ASD-specific level 1screening tools for children aged,36months. These include the followingtools.

CHAT

The CHAT was the first ASD screeningtool to be assessed at a populationlevel.46 It cannot be recommended,however, for current early detectionefforts due to its low sensitivity (18%,based on 6-year follow-up of a screenedcohort of 18-month-olds).49

Q-CHAT

The Q-CHAT extends the measurementmodel of the CHAT, covering a broaderrange of ASD symptoms, which arerated on a 5-point scale (rather thanpresent/absent). Preliminary datasuggest that the Q-CHAT distinguisheschildren with ASD from low-risk 18- to24-month-olds.33 A recent secondaryanalysis using the 10 Q-CHAT items thatbest discriminated groups with andwithout ASD and that optimized ascreening cut-point indicated sensitiv-ity and specificity estimates as high as91% and 89%, respectively, in a case-control sample.34 Further validation ofthis abbreviated screen is needed,however, in independent, community-based samples similar to where thescreen would be used.

M-CHAT

TheM-CHAT, alsoadapted from theCHAT,has been assessed in large community

SUPPLEMENT ARTICLE

PEDIATRICS Volume 136, Supplement 1, October 2015 S43 by guest on May 8, 2020www.aappublications.org/newsDownloaded from

TABLE1

Parent-ReportScreeningToolsforAutism

ScreeningTool

Reference

Population(N,Age,Diagnosis,Level)

SensitivityandSpecificity

PPVandNPV

Comments/Recom

mendation

ASD-specificscreeners:parent

report

FYI

Reznicketal,20

2007

(Turner-Brow

netal,21

2013)

N=698infantsaged

12mo

Preliminaryfindings:

(PPV

=0.31

andNPV=0.99)

Prom

isingtoolforinfantsaged

12mo,butadditionaldataneeded

Generalpopulationmailingat12

mo,

with

follow-upat

42mo

N=699with

outcom

esat42

mo.

Diagnosiswith

ASD=9.FYI2-dom

ain

risk

algorithmflagged4/9caseslater

diagnosedwith

ASD

Sensitivity=0.44;Specificity=0.99

ESAT

Dietzetal,22

2006;

Swinkelsetal,23

2006

N=31

724from

generalpopulation

Sensitivityandspecificitynotreported

PPV=0.25

Notyetrecom

mendedas

level1

screener;additionaldataneeded

Stage1,n=370screened

positive;

Identified

18ASDfrom

31724screened

Stage2,ofn=255100screened

positive

14–15

mo(m

ean:14.9mo)

M-CHAT

Robins

etal,24

2001

N=1293,m

ixoflowandhigh

risk

Estim

ates

ofsensitivityandspecificity

cannotbe

determ

ined

from

thisstudy

(screen-negativecasesnot

system

aticallyevaluated)

Strong

evidence

foru

seas

bothlevel

1andlevel2

tool,16–30

mo;

additionaldatawillbe

helpful,

especiallyinestim

ating

sensitivity

Mean:14.9mo(14–15

mo)

Robins,25

2008

N=4797;362

screened

positive

(qualified

forfollow-upinterview);

16–26

mo

Estim

ates


cannotbe

determ

ined

from

thisstudy


system

aticallyevaluated)

Withoutfollow-upinterview,PPV

=0.058

15-,18-,and24-mowell-childvisitresults

With

follow-upinterview,PPV

=0.57

Ifscreen-positive

casesareexam

ined

foranysignificant

developm

ental

delay,thePPVforM-CHAT+follow-up

interviewis.0.90

across

studies

Kleinm

anetal,26

2008

n=3309

lowrisk,n

=484high

risk

Estim

ates


cannotbe

determ

ined

from

thisstudy


system

aticallyevaluated)

PPV=0.11

forlow-risksample,0.60

for

high-risksample,withoutinterview

.Thisimproved

to0.65

(low

risk)and

0.76

(highrisk)whenfollow-up

interviewwas

considered

partofthe

screeningprocedure

16–30

mo

Pandey

etal,27

2008

n=6050

lowrisk

andn=726high

risk

Estim

ates


cannotbe

determ

ined

from

thisstudy


system

aticallyevaluated)

PPV=0.43

forlow

-risksamples

(younger

andoldercom

binedforthistable)and

0.76

forhigh-risksamples;PPV

calculated

basedon

M-CHAT+follow-

upinterview

16–30

mo

Note:Sam

pleoverlaps

with

Kleinm

anetal,26

2008,but

does

notinclude

samples

from

Robins

etal,24

2001,or

Robins,25

2008

Inadaetal,28

2011

N=659;18

mo

Estim

ates


cannotbe

determ

ined

from

thisstudy

(screennegativecasesnot

system

aticallyevaluated)

PPV=0.733


TABLE1

Continued

ScreeningTool

Reference



PPVandNPV

Comments/Recom

mendation

Canal-Bediaetal,29

2011

Validity

study:2417

lowrisk

and63

high

risk;18–36

mo

Estim

ates


cannotbe

determ

ined

from

thisstudy

(screennegativecasesnot

system

aticallyevaluated)

Validity

study:PPVnotreported

separatelyforlow-riskandhigh-risk

samples

Invalidity

study,19

of23

children

diagnosedwith

ASDwerefrom

high-risksample.

Reliabilitystudy:2055

lowrisk;18–36

mo

Reliabilitystudy:PPV=0.19

Inreliabilitystudy,rateofASDinlow-

risk

samplewas

2.9in1000

Pinto-Martin

etal,30

2008

N=152;18–30

mo

Nodiagnosticfollow-up,cannotassess

psychometrics;com

parisonofM-CHAT

andPEDS

Chlebowskietal,3

12013

N=18

989,18–30

mo

Estim

ates


cannotbe

determ

ined

from

thisstudy

Amongscreen-positive

childrenwho

wereevaluated(171

of278[60.7%

])Em

phasizes

potentialclinicalutility

ofM-CHATas

level1

screen

(high

PPV)

Note:som

epotentialfalse-negative

findings

ascertainedby

concurrent

screeningusingotherinstruments

PPV=0.538forASD

(ifany

DDisincluded,

PPVincreasesto0.977)

AuthorssuggestedthatifinitialM-

CHATscoreis$7,thefollow-upM-

CHAT

interviewmay

notb

eneeded

duetohigh

PPVforASD

(.0.80).However,the

follow-up

M-CHATinterviewisessentialfor

childrenwith

initialscores

of3–6

Note:Sam

pleoverlaps

with

Kleinm

anetal,26

2008;Pandeyetal,27

2008;

Robins

etal,24

2001;Robinsetal,25

2008

M-CHAT-R/F

Robins

etal,32

2014

N=16

115low-risktoddlers

Estim

ates


cannotbe

determ

ined

from

thisstudy

Amongscreen-positive

childrenwho

wereevaluated(221

of348[63.5%

])Childrenwith

,3itemsendorsed

(93%

ofallcases)didnotrequire

thefollow-upinterviewor

any

otherevaluation.Childrenwith

3–7itemsendorsed

(6%ofall

cases)required

thefollow-up

interview;ifatleast2items

remainedpositive,then

referral

fordiagnosticevaluationwas

indicated.Childrenwith

$8items

endorsed

(1%ofallcases)were

atsufficiently

high

risk

tobe

referred

directlyfordiagnostic

assessment.Usingthisstrategy

reducedthecase

positiverate

(from9.2%

to7.2%

)without

significantchange

toPPV,relative

toprevious

follow-upM-CHAT

strategy

Note:som

epotentialfalse-negative

findings

ascertainedby

concurrent

screeningusingotherinstruments

PPV=0.475forASD

(ifany

DDisincluded,

PPVincreasesto0.946)

Q-CHAT

Allison

etal,33

2008

779low-risktoddlerswith

meanageof21

mo;plus

Sensitivityandspecificitynotp

rovided

Notreported

160toddlers

andpreschoolers

with

ASD

with

meanageof44

mo

SUPPLEMENT ARTICLE


TABLE1

Continued

ScreeningTool

Reference



PPVandNPV

Comments/Recom

mendation

Allison

etal,34

2012

754controls;m

ean:36

mo(drawnfrom

low-risksampleinAllison

etal,332008)


PPV=0.58

(with

pretestodds=0.16

basedon

availablesample)

Clinicaldiagnosesbasedon

parent

report,w

ithrecruitm

entthrough

Web-based

research

registry

126toddlers

andpreschoolers

with

ASD

(aged15–47

mo;mean:20.8mo)

Basedon

screeningcut-point

of3from

derivationsample,usingthe10

of25

itemsfrom

originalQ-CHAT

thatbest

discriminategroups

Furtherevaluationinindependent

samples

warranted

Random

lyallocatedtoderivationand

validationsamples

DBC-ES

Gray

etal,35

2008

N=207;20–51

mo;level2

Sensitivity=0.83

(estimated);Specificity

=0.48

(estimated)

PDDR

SEavesand

Williams,362006

N=199with

autistic

disorder,rated

byteachers,teachinginterns,andfamily

mem

bers;aged1–6y

Factor

analysis,nopsychometrics

calculated

Insufficientdatatoevaluateutilityas

screeningtoolforyoungchildren

Eavesetal,37

2006

N=134,ratedby

teachers,teaching

interns,or

parents;aged

3–26

y(m

ean:9.7y);diagnosis:autism(n

=86),Asperger

disorder

(n=11),PDD-

NOS(n

=15),non–ASDdisorder

(n=

23)

Autistic

disorder,cutoff85:Sensitivity=

0.93

andspecificity=0.48

Notreported

Autistic

disorder,cutoff90:Sensitivity=

0.84

andspecificity=0.58

PDD,cutoff85:Sensitivity

=0.88

and

specificity=0.68

PDD,cutoff90:Sensitivity

=0.78

and

specificity=0.77

ASD-specificscreeners:interactiveobservationalm

easures

STAT

Stoneetal,38

2000

n=40

(developmentsample),n

=33

(validationsample);24–35

mo,level2

(highrisk)

Sensitivity=0.83andspecificity=0.86for

validationsample(sensitivity

=0.83

andspecificity=0.83

fordevelopm

ent

age-matched

subsam

ple)

Notreported

Strong

evidence

foruseas

level2

tool,24–35

mo;prom

isingfor14–

23mobutadditionaldatawillbe

helpful

Stoneetal,39

2004

Study1:N=52,24–35

mo,ASDandother

developm

entaldelay

matched

onchronologicaland

mentalage,level2

Study1:one-halfofsampleused

todeterm

inecutoffwith

optim

alsensitivity/specificityandone-half

used

tovalidatecutoffof2:Sensitivity

=0.92

andspecificity=0.85

Study1:PPV=0.86

andNPV=0.92

(validationsubsam

ple)

Study2:N=104,24–35

mo,level2

Study2:notreported,butbased

ontable

provided,sensitivity

=1.0and

specificity=0.90

forautistic

disorder

(low

erforPDD-NO

S)Stoneetal,40

2008

N=71,12–23

mo,level2

(follow-up

assessment24–42

mo)

Cutoffof2:Sensitivity=1.0andspecificity

=0.40

Cutoffof2:PPV=0.38

andNPV=1.0

Cutoffof2.75:Sensitivity

=0.95

and

specificity=0.73

Cutoffof2.75:PPV

=0.56

andNPV=0.97

Cutoffof2.75

insubsam

pleofchildren

14–23

mo(n

=50):Sensitivity=0.93

andspecificity=0.83

Cutoffof2.75

insubsam

pleofchildren

14–23

mo(n

=50):PPV=0.68

and

NPV=0.97


TABLE1

Continued

ScreeningTool

Reference



PPVandNPV

Comments/Recom

mendation

BISCUIT

Matsonetal,41

,422009

N=1007

samplewith

ASDor

DDaged

17–37

mo(m

ean:26.4mo)

330confirm

edASDdiagnosis

Prom

isingas

diagnostictoolor

level

2screener;m

oredataneeded


AUC=0.55

SORF

Wetherbyetal,43

2004

N=150,level2

screen

oflow-risk

sampleof6581

childrenaged

18–24

mo

20Significant

redflags;cutoffof8red

flags

Notreported

Prom

isingas

level2

screener;

reported

databasedon

coding

from

videorather

than

office-

basedassessment;moredata

needed

Sensitivity=0.87

andSpecificity=0.84

AUC=0.93

Broadbandscreener:parentreport

CSBS

ITC

Wetherbyetal,43

2004;

Wetherbyetal,44

2008

N=5385

childrenaged

6–24

mo

screened;follow-upevaluationofn=

813;generalpopulation(consecutive

screensfrom

6to24

mo;diagnostic

outcom

e/follow-upquestionnaire

at4

+years)

60ASDfrom

5385

screened

Note:PPV

andNPVarenotreported

for

ASDspecifically

butfor

alldelays

collapsed

into1group

Need

additionaldata,butpromising

toolfor9–.24

mo;not

recommendedforage,9mo

Mean:16.4mo(6–24

mo)

PPVandNPVforcommunicationdelay

vary

with

age:

Note:ITC

positivescreen

does

not

distinguishcommunicationdelay

from

ASD;second-levelASD-

specificscreen

recommended

Sensitivity=88.9–94.4;Specificity=88.9

PPV=0.42–0.79

Sensitivityestim

ated

at0.93

(56/60

identified

when,6consecutive

screensweredone

across

18-mo

period)

NPV=0.87–0.99

Pierce

etal,17

2011

N=10

479infantswhose

parents

completed

checklistat1-ywell-child

visit(meanage:12.54mo);level1

screen

32with

ASDfrom

10479screened

PPV=0.75(estimated)fordevelopm

ental

delays

includingASD

Prom

isingtool;furtherstudyneeded

toexam

inesensitivityand

specificityingeneralpediatric

settings

Sensitivityandspecificitynotprovided

Oosterlingetal,45

2009

N=238;8–44

mo(m

ean:29.6mo)

Seenote(atright)

Seenote(atright)

Note:Truesensitivity,specificity,PPV,

andNPVcannotbe

estim

ated

becauseanothertool(ESAT)was

used

as“prescreen”andonly

ESAT

screen-positive

caseswere

subsequentlyscreened

with

ITC

Overallsensitivity

reported

at0.71

OverallPPV

reported

at0.78

Overallspecificityreported

at0.59

OverallN

PVreported

at0.50

Becausenorm

sforITCareavailableonly

forchildrenaged

6–24

mo,itis

difficulttointerpretuse

asascreener

forchildrenaged

.24

mo

Adaptedfrom

thetabledevelopedby

theAutismSubcom

mittee

fortheNationalChildren’sStudy.Otherinstrumentsnotlistedinthetable:CHAT

(Baron-Cohen

etal,46

1992;Baron-Cohen

etal,47

1996;and

otherarticles);longitudinaldataindicatedpoor

sensitivity;notrecommendedforuseas

level1

screen;and

SocialResponsiveness

Scale–Preschool(SRS-P),under

developm

entforscreeningpreschool-agedchildren.ADI-R,AutismDiagnosticInterview–Revised;AOSI,AutismObservationScalefor

Infants;BISCUIT,Baby

andInfant

Screen

forChildrenwith

Autism

Traits;CSBSITC,CommunicationandSymbolic

Behavior

Scales

Infant

ToddlerChecklist(orInfant

ToddlerChecklist,ITC);DBC-ES,Developm

entalB

ehaviorChecklist–EarlyScreen;DD,

developm

entaldelay;ESAC,EarlyScreeningforAutism

andCommunicationDisorders;ESAT,EarlyScreeningofAutistic

Traits;M

-CHAT-R/F,M-CHAT,RevisedWith

Follow-Up;PDD-NO

S,pervasivedevelopm

entaldisordernototherw

isespecified;PDD

RS,

PervasiveDevelopm

entalD

isorderRatingScale;Q-CHAT,Quantitative-ChecklistforAutisminToddlers;SORF,System

aticObservationofRedFlagsofASD;STAT,Screening

ToolforAutisminToddlers

&YoungChildren.

SUPPLEMENT ARTICLE


samplesasa level 1screen. The23-itemM-CHAT questionnaire, combined witha follow-up interview to help clarifyitems endorsed by parents on theinitial screen, is estimated to havea PPV as high as 0.57 to 0.65 in low-risksamples.25,26,31 Pandey et al27 reportedthat the PPV of the M-CHAT (as used forfirst-level screening in a low-riskcommunity sample with follow-up in-terview) is lower in younger children,with a PPV of 0.28 in toddlers aged 16to 23 months compared with a PPV of0.61 in those aged 24 to 30 months.There are many reasons for false-positive findings, including develop-mental concerns that may resolve andbehaviors in typically developing tod-dlers that overlap with ASD deficits,such as repetitive behaviors (eg,turning lights on and off) and re-stricted interests (eg, insistence onroutines).19 However, despite lowerspecificity for autism at 18 months,PPV for any diagnosable develop-mental disorder was high for allgroups. In the largest sample of tod-dlers (aged 18–30 months) reportedto date (N = 18 989 [including somechildren in previous reports]),25–27 thePPV of the M-CHAT for ASD was 0.54,and for any developmental disorder, itwas 0.98.31 As in other community-based ASD-screening studies, esti-mates of PPV were based on thosescreen-positive children who attendedand completed a diagnostic evaluation(39.3% of screen-positive childrenwere not assessed).

The M-CHAT has also been evaluatedinternationally and in multiple lan-guages. Canal-Bedia et al29 assessedthe reliability and predictive validity ofa Spanish translation of the M-CHAT ina combined community and at-risksample in Spain. The PPV in the com-munity sample was 0.19, although thisfinding may have reflected a relativelylow base rate of identified preschool-aged children with the disorder (2.9 in

1000) (Table 1). Another study thatevaluated the psychometric proper-ties of the Spanish version of theM-CHAT in a community sample ofchildren in Mexico reported similardiscriminative validity,50 althoughsome items appeared less informativefor ASD than in published reports onthe original English-language version.Psychometric data on Japanese28 andArabic51 translations have also beenreported. (Additional information onavailable translations of the M-CHATis available at http://www2.gsu.edu/∼psydlr/Site/Official_M-CHAT_Website.html [accessed October 17, 2014]).

Recently, Robins et al32 reported vali-dation data for a new version of thisscreening tool, the M-CHAT, Revisedwith Follow-Up, in 16 115 toddlers. Thequestionnaire was reduced to 20items, removing 3 items that hadperformed poorly (“peek-a-boo,”“playing with toys,” and “wanderingwithout purpose”); wording on otheritems was simplified and/or examplesprovided for further clarity. A scoringalgorithm with 3 risk ranges was de-veloped. Children in the low-risk range(ie, ,3 items endorsed) did not re-quire the follow-up interview or anyother additional evaluation (93% of allcases). Children in the medium-riskrange (ie, 3–7 items endorsed [6% ofall cases]) required the follow-up in-terview to clarify their risk for ASD; if atleast 2 items remained positive, thenreferral for diagnostic evaluation wasindicated. Children in the high-risk range(ie,$8 items endorsed [1% of all cases])were at sufficiently high risk to be re-ferred directly for diagnostic assess-ment without the follow-up interview.This revised scoring and referral algo-rithm reduced the initial screen-positiverate (from 9.2% to 7.2%) and increasedthe overall rate of ASD detection (67 vs 45per 10 000) compared with the originalfollow-up M-CHAT.

Early Screening for Autistic Traits

Population screening at an even earlierage has been associated with higherfalse-negative rates (lower sensitivity),which is somewhat expected given theslow onset of symptoms that emergesacross the first 24 months of life. TheESAT was assessed in a large (N = 31724) population sample of 14- to 15-month-olds, with a low case detectionrate (,1 in 1000).22,23 Moreover, PPV ofthe ESAT was only 0.25, which wouldpotentially lead to the referral ofa large number of toddlers without ASDbased on a positive screen (PPV forother developmental delays was notreported). The authors recommendeda second screening at 24months of ageto identify children who regress afterage 18 months or those who aremissed for other reasons.

Baby and Infant Screen for ChildrenWith Autism Traits

Preliminary data on the Baby and InfantScreen for Children with Autism Traitstool indicate good discriminationbetween toddlers with known ASDdiagnoses and those with other de-velopmental delays as identified clini-cally.41 Additional data are needed,however, to confirm how this measurewould perform in a screening context.

FYI

The FYI is a parent questionnairedesigned to screen for signs of autism in12-month-olds. Initial data on the FYIsuggest the potential for modest sensi-tivity.20 In a recent prospective follow-upstudy of a community sample of 699children whose parents initially com-pleted the FYI at approximately thechild’s first birthday, 4 of 9 childrensubsequently diagnosed with ASD at 3years of age were identified. A scoringalgorithm that optimized prediction ofASD identified 13 (1.9%) of 699 partic-ipants who met cutoffs on 2 domains(social communication and sensory


http://www2.gsu.edu/~psydlr/Site/Official_M-CHAT_Website.html



regulation).21 Assessment of PPV in anindependent/validation sample is stillneeded.

The working group suggested that ad-ditional efforts are needed to developand validate population-based ASDscreening tools aimed at the 12- to 18-month age range, anticipating thatmodest sensitivity at this age may war-rant follow-up with additional screeningat a later age (eg, at 24 months). In ad-dition, the working group recommendedthat standardized screening specificallyfor ASD should be performed whenparents raise concerns between well-child visits or when concerns areraised upon general developmentalsurveillance or screening duringscheduled visits. Parental concern ef-fectivelyraises thepriorprobability thata childwill have ASD, thereby increasingthePPVof a screening test regardlessofits intrinsic sensitivity and specificity.

Level 2 Screening Tools

Two interactive observational assess-ments have been developed for use aslevel 2 screeners in young childrenidentified as being at high risk of ASD.

Screening Tool for Autism inTwo-Year-Olds

The Screening Tool for Autism in Two-Year-Olds (STAT) has been assessed inclinical samples of 2-year-olds referredfor suspected ASD, with a sensitivity andspecificity as high as 92% and 85%, re-spectively.39 Recent data indicate thatthe STATmay also have utility in youngertoddlers aged 14 to 23months, althoughadditional data are needed for this agegroup.40 Although the STAT requiresa higher level of expertise to administerthan parent questionnaires such as theM-CHAT, a recent study provided evi-dence of the effectiveness of Web-basedtraining of community services pro-viders of various professional back-grounds; this training could enhancethe feasibility of the STAT.52

Systematic Observation for Red Flags

The Systematic Observation for RedFlags has shown promise in discrimi-nating ASD from other communicationdelays.43 Additional data are needed ina screening context.

Broadband screening in childrenaged ,24 months can assist in earlydetection of ASD

Delays and deviances in social commu-nication are often subtly present aroundthefirstbirthdaybutareoftennotstronglyASD-specific at that early age. Broadbanddevelopmental screening tools, such asthe Communication and Symbolic Be-havior Scales Developmental Profile(CSBS DP) Infant/Toddler Checklist de-veloped by Wetherby and Prizant,53 wereshown to be effective at detecting autismbefore the onset of full-blown clinicalsymptoms. Wetherby et al44 evaluatedthe CSBS DP Infant/Toddler Checklist ina community sample of 5385 childrenaged 6 to 24 months recruited fromhealth and child care services. TheInfant/Toddler Checklist identified 56(93%) of 60 children with ASD classifiedindependently at age 3 years in a con-current prevalence study of the sameregion. Some Infant/Toddler Checklistfindings were positive as early as 9 to 11months, although in some cases, an ini-tial screen was negative at 9 to 11months and did not becomepositive untila lateradministration. The Infant/ToddlerChecklist also identified concerns soonerand more consistently than an open-ended question about parents’ de-velopmental concerns. Subsequently,Pierce et al17 assembled a network of137 pediatricians who administered theCSBS DP Infant/Toddler Checklist at ev-ery routine 1-year check-up examination.Of ∼10 000 screens administered, 1318children failed the screen. The pedia-tricians referred 346 screen-positivechildren as “at-risk” children (thescreening was thus embedded withina surveillance context, in which clinicaljudgment contributed to referral deci-

sions); 184 ultimately received furtherevaluation. Of this group, 32 toddlers re-ceived an ASD diagnosis by age 3 years.This general population screening ap-proach also detected 65 toddlers witha language delay or global developmentaldelay, and 36 children with other delays.Thus, the PPV for detecting toddlers withASD or developmental delay in this studywas estimated to be 0.75. Importantly, alltoddlers identified with delays were re-ferred for treatment, and the majoritystarted intervention well before theirsecond birthday.

This research illustrates that autism cansometimes be detected by the firstbirthday by using a broadband de-velopmental screen in real-world pedi-atric practices as standard of care. TheCSBS DP Infant/Toddler Checklist is notspecific forASD (ie, doesnot differentiateASD from other communication dis-orders), but follow-up evaluation bya developmental specialist (eg, speechlanguage pathologist, psychologist, de-velopmental behavioral pediatrician)can help determine the need for ASD-specific diagnostic assessment as wellas identify other developmental delays inneed of support and intervention. Useof even broader, more general de-velopmental screening tools, such as theParents’ Evaluation of DevelopmentalStatus (PEDS)54,55 and the Ages & StagesQuestionnaire,56 to detect ASD are underinvestigation. Because these tools arecommonly used in pediatric practice, itwill be important to determine theirutility in detecting ASD in the secondyear of life even though their sensitivityand specificity are not expected to be ashigh as those of ASD-specific screeners.

Statement 2: The evidenceindicates that siblings of childrenwith ASD are at elevated risk forASD and other developmentaldisorders and thus should receiveintensified surveillance.

Basedondata fromaUSregisterof2920children aged 4 to 18 years in families

SUPPLEMENT ARTICLE


affected by ASD, the frequency of ASD ina later-born sibling has been estimated at14%.57More recently, several independentgroups conducting prospective longitudi-nal research involving infant siblings ofchildren with ASD reported a pooled es-timated recurrence risk of 18%.58 In con-trast, a recent population registry-basedstudy from Denmark59 estimated re-currence risk at closer to the 7% to 8%level reported in older studies.60 Regard-less, rates of ASD in siblings greatly ex-ceed population risk, emphasizing theneed for intensified monitoring. More-over, younger siblings of children withASD demonstrate significant deficits onindices of social communicative develop-ment and cognitive functioning, aswell aselevated ASD symptoms relative to youn-ger siblings of typically developing chil-dren.61–64 Because these children are atelevated risk, they require intensifieddevelopmental surveillance. At a mini-mum, they should receive continuoussurveillance for developmental issuesand be screened for ASD at 18 and 24months of age, as recommended by theAAP for all children.2

Statement 3: Children identifiedthrough ASD-specific screeningshould be immediately referred fordiagnostic/developmentalevaluation and appropriateintervention.

The AAPhas recommended that childrenwho screen positive on an ASD-specificscreening tool be scheduled for a com-prehensive evaluation and referredconcurrently to early intervention ser-vices as appropriate.2 Available inter-ventions are mandated in the UnitedStates but vary in availability and qualityby locality, and theymay consist of non–ASD-specific public early interventionprograms, such as speech therapy, andearly childhood education programs.

It is hoped that early screening will leadto improved outcomes as a result ofearlier referral and earlier initiation ofintervention. However, recent studies

suggest that such benefits of earlyscreening frequently go unrealized. In anational study of 17 pediatric practices,implementation of general develop-mental screening did not always lead toreferral of screen-positive children toa medical subspecialist or early in-tervention programs.12 These inves-tigators noted that some families didnot understand the reason for a follow-up evaluation. Additional research isneeded to address how to better engagefamilies in the screening process to fa-cilitate rapid follow-up, as well as toidentify and characterize other potentialbarriers to early diagnosis and treat-ment related to system capacity orprovider attitudes and practices.

Statement 4: The long-termstability of ASD diagnosis inchildren aged ‡24 months is wellestablished. Emerging datasuggest that ASD diagnoses insubstantial proportions of childrendiagnosed before age 24 monthsare also stable, although furtherresearch is needed, particularly inthe context of early screening.

Ten articles were identified in whichchildren received an initial diagnosticassessment forpossibleASDbeforeage3 years and were then reassessed atleast 1 year later.65–74 In general, thestability of ASD diagnoses establishedat $24 months (ie, the rate at whichan ASD diagnosis was confirmed onreassessment) was very high, rangingfrom 68.4% to 100% when the initialdiagnosis was autistic disorder (me-dian: 92%), and from 40% to 100%when the initial diagnosis was perva-sive developmental disorder not oth-erwise specified (according to theDiagnostic and Statistical Manual ofMental Disorders, Fourth Edition, orthe Diagnostic and Statistical Manualof Mental Disorders, Fourth Edition,Text Revision [median: 61%]).

Fourof these studies involved samples ofchildren aged,24months (Table 2),65–68

although only 1 study focused almostexclusively on this age group.67 Thesestudies provide promising evidence ofthe stability of ASD diagnosed as early as14 months; the samples were relativelysmall, however, and there is no directcomparison of stability in children di-agnosed before versus after age 24months. Of note, 2 studies focused ontoddlers identified by using community-level ASD screening before age 24months.65,66 Both studies indicated highdiagnostic stability for children initiallydiagnosed with autistic disorder (85%–93%) but more modest stability forchildren diagnosed with pervasive de-velopmental disorder not otherwisespecified (47%–62%). Further researchin larger samples is needed, but theevidence to date supports the stability ofASD diagnoses before age 2 years.

Statement 5: Further attention topotential barriers to ASD-specificscreening in the health caresystem is needed.

Pediatricians have noted major barriersto screening, including the following: lackof time and inadequate reimbursement;logistic challenges, such as disruption ofwork flow, lack of familiarity with tools,and difficulty with scoring; and lack ofoffice-based systems for making refer-rals and monitoring outcomes.

Lack of Time and Reimbursement

Insufficient time and inadequate re-imbursement are often cited by pro-viders as barriers to performingscreening.12,13,19,75 Pediatricians havea limited amount of time to complete anincreasing number of tasks, includingscreening for non-ASD disorders, duringa well-child visit.19 Selection of a broad-band screening instrument would meetwith greater acceptance if the tool coulddetect multiple developmental dis-orders of interest. Busy periods, such asthe onset of the winter viral season, of-ten impede the ability of a practiceto consistently screen.12 To optimize


screening, some practices have in-stituted ongoing data collection andmonitoring of their efforts.

The lack of reimbursement for screeningiscommonlycitedasabarrier.However, in1 study, the 3 practices that routinelyscreened at the 30-month well-child visitreported no difficulties in collecting pay-ment.12 In another study,17 pediatric offi-ces received no payment at all forscreening but rather received trainingand data collection support, as well asstreamlined follow-up diagnostic assess-ments for screen-positive children. Thus,reimbursement challenges may be me-diated by infrastructure support (eg, stafftraining/mentoring) to make screeningeasier to implement, as well as timelyaccess to appropriate follow-up. In thisway, pediatricians may be reassured thatthere is capacity in the health system tosupport children who screen positive.

Logistic Challenges

Other challenges to screening imple-mentation include concerns over a dis-

ruption of work flow, unfamiliarity withscreening instruments, and difficulty withscoring.12,13,75 Providers often expressconcerns about how to distributescreeningquestionnaireswithout slowingthe flow of patients through the office.12,13

Nevertheless, in a national sample of 17pediatric practices, .85% of childrenpresenting at recommended screeningages were screened, with practices di-viding responsibilities among staffmembers and proactively monitoringimplementation.12 Miller et al76 found thatscreening at sick visits was necessary toachieve coverage of the age-eligible chil-dren, especially for the small number ofuninsured children. Training of office staffas well as professional education canremedy a lack of familiarity with the useand scoring of screening tools.

Lack of Office-Based Systems forMaking Referrals and MonitoringOutcomes

In the sample of 17 pediatric practices,only 61% of children with failed screens

were referred, and many practicesstruggled to track their referrals.12

Practice-specific referral rates variedwidely, from 27% to 100%. It is importantthat each pediatric practice establisha specific implementation system to ex-pedite referrals, communicate with spe-cialists and early intervention programs,and track follow-through and outcomes.Clearly, early screening initiatives areonly as effective as access to resourcesfor follow-up evaluation and early in-tervention. Communication back to thereferring office relative to the outcomesof follow-up actions is critical if only toreassure all concerned of the value ofsuch referrals. For children with ASDs,early intervention services have becomemore accessible through Part C of the2009 Individuals With Disabilities Educa-tion Act but access may not be equal inall parts of the country, and the quality ofservices can vary widely and affect out-comes. Indeed, although the NationalResearch Council has recommendedentry into an intervention program as

TABLE 2 Studies of Diagnostic Stability That Include Children Initially Assessed With ASD Before 2 Years of Age

Reference Sample Mean Age, AgeRange at T1, mo

Mean Age, AgeRange at T2, mo

Diagnosisat T1

N Diagnosis at T2 N % Stability

van Daalen et al,65 2009 Population-based sample 23 43 (34–64) Autism 40 ASD 38 95.0Non-ASD 2

PDD-NOS 13 ASD 8 61.5Non-ASD 5

Non-ASD 78 ASD 2 97.4Non-ASD 76

Kleinman et al,66 2008 Mixed level 1 (physician office)and level 2 (early intervention,sibling) sample

26.7 (16–35) 52.9 (41–82) Autism 46 ASD 39 84.8Non-ASD 7

PDD-NOS 15 ASD 7 46.7Non-ASD 8

Non-ASD 16 ASD 0 100Non-ASD 16

Chawarska et al,67 2007 Referrals to specialty clinicwith suspected ASD

21.6 (14–25) 35.9 Autism 21 ASD 21 100Non-ASD 0

PDD-NOS 6 ASD 6 100Non-ASD 0


Gillberg et al,68 1990a Referred sample 23.0 (8–35) 57.7 (36–140) Autism 21 ASD 21 100Non-ASD 0

PDD-NOS 4 ASD 2 50Non-ASD 2


PDD-NOS, pervasive developmental disorder not otherwise specified; T1, initial diagnostic assessment of ASD; T2, reassessment of ASD diagnosis, at least 1 year later in these studies.a One child, diagnosed at 8 months, was followed up only to age 26 months and thus was excluded from the table.

SUPPLEMENT ARTICLE


soon as an ASD is suspected,3 localfactors, including funding, can affectaccess to services (wait-listing) ormakecertain early intervention programsunavailable to some children.75,77

Thus, barriers to screening can be over-come with specific strategies such astrainingand involvementof clinic staff anduse of reminder systems, even in busypractices. However, better-coordinatedefforts are needed to ensure access tospecialized assessment and interventionfor children at risk identified through thescreening process, as well as communi-cationback tocommunitypediatricians. Inaddition, further consideration is neededregardinghowphysicianbeliefs related toASD screening (eg, potential risks andbenefits to children and families, systemcapacity to provide timely specialized as-sessment and treatment services) mayinfluence practice behavior. Such beliefscan contribute to incongruence betweenphysician knowledge and actions whenmanaging ASD-related concerns78 andthus may also need to be addressed tofacilitate uptake of ASD screening intocommunity pediatric practices.

Statement 6: Methodologicallyrigorous research in ASD-specificscreening should be a high priority.

Futureresearch inASDscreeningwouldbe aided by attention to the followingmethodologic issues:

� use of large, representative high-and low-risk samples, to strengthenthe generalizability of findings

� use of meaningful end points (eg,validated diagnostic measures toassess for ASD and other develop-mental disorders, as well as anincreased focus on outcomes ofgreatest relevance to families andto the health system, such as age ofdiagnosis, age of entry into inter-vention, and long-term developmen-tal gains resulting from screening)

� inclusion of systematic surveil-lance methods, as well as follow-

up tracking of screen-negativecases, to improve estimates of sen-sitivity, specificity, and NPV

� evaluation of different scoringapproaches (categorical versuscontinuous) and, potentially, differ-ent age-specific scoring algorithmsfor specific ages, to further optimizescreening strategies that might beimplemented longitudinally

� reporting of detailed characteriza-tions of study participants, includ-ing social factors, cognitive level,and medical history, to improvecomparisons across studies andto better understand what factorsmight influence the accuracy ofscreening for individual children

� evaluation of potential differencesbetween screen-positive childrenwho are seen for a diagnostic as-sessment and those who do notcomplete follow-up (which is oftenin the range of 25%–40%25,27 and insome studies exceeds 50%17) tofurther evaluate potential barriersand facilitators, and provide infor-mation essential to evaluating thegeneralizability of study findings

� inclusion of underrepresented mi-nority and historically underservedgroups, to help ensure representa-tive samples and the development ofculturally appropriate adaptations ofscreening tools for such populations

Lower socioeconomic status and non-white ethnicity (particularly Hispanic)have been associated with delayed ageof diagnosis, potentially due to dis-parities in access tohealth services.79–81

However, there is evidence that appli-cation of standardized screening canhelp reduce such disparities and en-sure timely diagnosis of children acrossa diversity of backgrounds.82

Statement 7: Additional priorities forfuture research include studies that:

� Examine how broadband and ASD-specific screening tools can be used

in a complementary fashion to max-imize both sensitivity and specificityof early screening, perhaps in thecontext of multistage screening, inwhich a wide net is cast initiallyand false-positives are winnowedout in successive assessments

� Evaluate screening strategies by us-ing randomized experimental designs

� Consider additional outcome met-rics for screening: potential finan-cial savings to society, unintendedeffects (eg, family stress)

� Examine whether computer technol-ogy can improve screening accuracy

� Examine the effectiveness of re-peated screening for ASD

� Evaluate how belief systems affectscreening uptake and outcomes

� Examine potential screening strat-egies that include measurement ofbiomarkers

Examine how broadband andASD-specific screening tools can beused in a complementary fashion tomaximize both sensitivity andspecificity of early screening

Can a general developmental tool berelied on to identify children who shouldbe evaluated for ASD? If a broadbandscreening tool is indeed dependable, assuggested by Wetherby et al43,44 andPierce et al,17 then a multistage screen-ing strategy focusing on routine sur-veillance and use of a broadbandscreening tool, followed by an ASD-specific instrument for children whotest positive on the initial screen, canhelp reduce the need for extra testingand the additional clinic time and effort.A notable value of this approach is thelimiting of referrals for specialized as-sessment, without sacrificing case de-tection rate. If broadband screeningcannot reliably detect ASD, then ascreening strategy mandating ASD-specific screening for all children,alongside broadband screening to


detect other potential developmentalconcerns,wouldbemoreappropriate. Thefirst approach was described by Filipeket al83; the second approach is currentlyrecommended by the AAP.2 Unfortunately,the effectiveness (and cost-effectiveness)of the 2 strategies has not been wellstudied. Data from a single pediatricpractice showed that ∼75% of childrenwith positive results on the ASD-specificscreening tool (the M-CHAT) were missedby the PEDS, a standardized general de-velopmental screening questionnaire.30 Itshould be noted, however, that this studydid not report actual ASD diagnoses butrather simply examined agreement inscreening classification by the 2 tools.However, Wiggins et al54 reported that theM-CHAT had higher sensitivity for ASD thanthe high-risk threshold for any area ofgeneral concern covered by the PEDS. Al-though the PEDS detected many childrenwith other developmental concerns, sen-sitivity for ASD could not be achievedwithout lowering the screen-positivethreshold to a level that would identifya substantial proportion of the generalpopulation (25%).

A study assessing the efficacy of sucha multistage screening program wouldalso assess/validate the effectivenessof: (1) training of health care pro-fessionals in recognizing early ASDsigns and using a specific screeningtool; (2) a specific referral protocol; and(3) feedback to the referring offices.

Evaluate screening strategies byusing randomized designs

The evaluation of ASD screening is oftenlimited to measurement of classificationaccuracy (estimates of sensitivity andspecificity, and/or PPV and NPV) withoutsufficient attention to whether the ulti-mate goals of screening are achieved(eg, earlier diagnosis and access totreatment) or the possibility that, as withother interventions, screening might beassociated with positive or adverse out-comes. Moreover, alternate approaches

to screening (eg, broadband versus ASD-specific, level 1 versus level 2, or somecombination) have never been directlycompared. We would argue that screen-ing is a public health intervention; that is,a comprehensive early detection strategyshouldnotbesolelybasedontheselectionof a particular screening instrument butrather must include other changes to theoverall system of care, such as enhancedtraining for health professionals and ex-panded capacity for early diagnosis andintervention by specialized teams. Thus,theoutcomesofscreeningmaynotsimplyberelatedtothemeasurementpropertiesof a tool but also to the successfulimplementation of other aspects to theoverall care pathway for children withsuspected ASD.17,84 As such, researchersshould explicitly define their screeningstrategy (ie, the screening instrumentplus collateral changes to the system ofcare) as well as the outcomes of interest,and evaluate the effectiveness of thesestrategies in real-life community settingsby using randomized designs. Random-ized designs have become the standardin other ASD intervention research (eg,Dawson et al5) and in other public healthscreening interventions.85 However, ob-servational studies will also need to becontinued because of the well-knownchallenges to constructing randomizeddesigns that reflect real-world clinicalpractice.86 Table 3 presents a comparisonof the relative strengths and limitationsof randomized and observational designswith respect to screening research.

Consider additional outcome metricsfor ASD screening

In the near term, evaluation of ASDscreening strategies will likely continueto focus on process measures, such asrates of targeted children screened, re-ferred, and diagnosed. However, ulti-mately, the idea of evaluating anyscreening program is to gauge its impacton distal health outcomes. For potentiallyfatal conditions, mortality is the ultimate

distal outcome. For nonfatal conditions,developing approaches to measure im-pact on morbidity, disability, or impair-ment can be a challenge. With respect toASD, although increases in referral andearly diagnosis rates can serve asmeaningful initial outcomes, screeningshould ultimately demonstrate a re-duction inpopulation impairmentandtheeffect of that impairment on society.Studies of ASD screening will thus even-tually need to consider the impact of thisscreening on long-term changes insymptoms and functional status. De-termining how to best measure thesedistal health outcomes is one of thechallenges of ASD research. In addition todistal health outcomes, assessing thecost impact of screening is often criticalto its eventual broad dissemination.

Because ASDs impose a sizable financialburden, not only in direct medical expen-dituresbutalsoinindirectcosts(eg,specialeducation services, lost productivity byfamily caregivers),87–89 a more in-depthunderstanding of these costs is neededto adequately comparedifferent screeningstrategies and to identify potential costsavings to society for those that are ef-fective. Finally, indirect costs associatedwith screening include an emotional di-mension. Evaluations of screening effec-tiveness, in addition to including distaloutcomes, need to consider these “costs”in addition to the financial costs associ-ated with false-positive findings.

Examine whether computertechnology can improve screeningaccuracy

The use of computer technology holdspromise for improving screening accu-racy. Parents can complete a screeningquestionnaire online and have accessto video exemplars for more accuratereporting. The capability to upload vid-eos can expedite specialist evaluations.A recent preliminary report suggestedthat the M-CHAT (including follow-upquestions) could be feasibly completed

SUPPLEMENT ARTICLE


electronically, with fewer false-positivefindings and good to excellent parentsatisfaction.90 Further studies should beconducted to determine feasibility andaccuracy in a larger sample of com-munity practices.

Examine the effectiveness of repeatscreening

ASD is heterogeneous in the pre-sentation and time course of core def-icits. Itwould thereforebe important fora screening program to administerASD-specific screening tools periodi-cally at differing ages to detect childrenat risk who, for a number of reasons,may have been missed on an earlieroccasion. Formal research can betterdefine the value and potential cost-benefit of repeat periodic screeningfor ASD, as well as identify potentialfactors that can improve the efficiencyand efficacy of specific approaches.

Examine how belief systems impactscreening uptake and outcomes

Belief systems of both providers andparents may influence screening out-comes. The uptake, or implementation,of clinical recommendations for screen-ingcanbediminishedifpediatriciansandother health care professionals havemisconceptions about ASDs (eg, a beliefthat children can “outgrow” ASD) or are

unfamiliar with pertinent inter-ventions.91 For example, cultural beliefsmay influence the significance attachedto differences in early social behavior orthe reporting of such differences tohealth care providers. A child who doesnotmake eye contactwith adults or pointmay not be worrisome if such behaviorsare considered disrespectful.19 Familiesmay also be less likely to participate infollow-up assessments92–94 if they arenot confident in the referring clinician’sskills and expertise.95 Studies examiningthe impact of belief systems would im-prove both provider and parental un-derstanding of diverse perspectives andinform targeted supports and inter-ventions.

Examine potential screeningstrategies that include measurementof biomarkers

Given that neuroanatomical abnor-malities in ASD have been shown tooccur consistently across develop-ment,96,97 and biological mechanisms(including genetic) may providea measurable “signature” even beforesymptom expression, there is hope thatspecific biomarkers may eventually beidentified that could contribute to earlydiagnosis. Indeed, recent studies fromdevelopmental neuroscience and mo-lecular biology have shown promise in

identifying specific markers that candistinguish children with ASD fromother high-risk and low-risk peers,even during infancy. However, most ofthese studies focused on group differ-ences rather than predicting outcomesat an individual level (needed to de-termine sensitivity and specificity)and/or focused on distinguishing chil-dren with known diagnostic statusrather than predicting diagnosis in chil-dren whose status is not yet known. Thissmall, yet growing body of researchincludes studies with well-defined high-risk cohorts (notably, younger siblings)aswell as general population cohorts thatbegin screening, tracking, and studyingthe biology of ASDs at 12 months. Whatboth approaches have in common is thatstudies are conducted within highly con-trolled research contexts. Thus, althoughbiomarker-based research holds consid-erable promise, the clinical utility ofincorporating such markers intocommunity-based early detection strat-egies remains to be demonstrated. Atpresent, no specific biomarkers arerecommended for ASD screening.

Several examples of studies using brain-based measures identifying candidatebiomarkers are summarized here to il-lustrate the potential contribution ofthis emerging field of research. Usingthe general population–based screening

TABLE 3 Designs for One-Step Evaluation Studies of ASD Screening Programs

Variable Randomized Trial Prospective Observational (Cohort) Retrospective Observational (Case-Control)

Comparison groups Young children randomly assigned todifferent screening control groups

Naturally occurring “screened” and“unscreened” groups in the community.Perhaps identified via different:

• Children with ASD having optimaloutcomes (“case subjects”)

• Health care providers • Age-matched children with ASD havingsuboptimal outcome representative ofthe population giving rise to the casesubjects (“control subjects”)

• Geographic area

Strengths • Random assignment to control forconfounding by indication

Temporality assured • Less cost

• Temporality assured • Less time to completeWeaknesses • Ethnically feasible? • Potential for selection bias • Potential for selection bias

• Large sample needed • Need to control for confoundingby indication

• Need to control for confounding byindication (more challenging inretrospective designs)

• Substantial follow-up needed • Large sample needed • Need to be able to accurately identifytrue screening encounter• Expensive • Substantial follow-up needed

• Can raise generalizability issues • Expensive


approach described by Pierce et al17 toassemble a cohort of toddlers with ASD,Dinstein et al98 recorded functional MRIactivity from 63 naturally sleeping tod-dlers with ASD, language disorder (ie,standardized score at least 1 SD belowthe mean), or typical development. Rel-ative to the other groups, toddlers withan ASD exhibited significantly weakerinterhemispheric correlations in the in-ferior frontal gyrus and superior tem-poral gyrus, 2 areas central to languageproduction and comprehension. Levels ofinterhemispheric coordination enabledaccurate identification of toddlers di-agnosed with ASD, with high sensitivity(72%) and specificity (84%). As anotherexample, Bosl et al,99 using the modifiedmultiscale entropy computed on the ba-sis of a resting state EEG, showed thatinfants at high risk for autism exhibita different developmental trajectory thantypically developing control subjects andthat these differences are most evidentbetween 9 and 12 months of age. Infantswere classified with.80% accuracy intocontrol groups and high-risk groups atage 9 months. More recently, Elsabbaghet al100 reported that evoked responsesto dynamic gaze at 8 months in high-riskinfants were predictive of an ASD di-agnosis at 36 months. In addition, Wolffet al101 described a pattern of bluntedwhitematter trajectories based on serialbrain MRI (using diffusion tensor imag-ing) between 6 and 24 months of age inhigh-risk infants with ASD symptoms at24 months; differences in these imagingindices were detectable by 12 months.

Blood-based biomarker studies of ASDhave yet to reveal themselves as viablescreening approaches,mainly due to thefact that discovered genetic mutationsoccur at relatively low rates in the ASD

population.Until recently, itwasreportedthat de novo genetic copy number var-iations are present only in 3% to 10% ofthe ASD population.102 However, recentdata using exome and whole genomesequencingmethods suggest the yield ofsuch testing for clinically informativevariants may be much higher.103,104

Moreover, although the contribution ofspecific biomarkers to risk predictionmay be modest, combined results froma panel of predisposing biomarkers canproduce information about an individu-al’s probability of developing ASD.105

Consideration of several biomarkers atonce is consistent with the multitude ofgenetic and epigenetic factors (and po-tentially other biological factors [eg,immune, indices of atypical braingrowth/connectivity]) that likely playa role in vulnerability to ASD in manychildren.106 The sensitivity and specific-ity for the risk score could be used toindicate the predictive performance ofthe biomarker combination. The ap-proach of combining multiple alleles/biomarkers to predict risk status hasalso been undertaken with other dis-orders of complex etiology, includingbreast and prostate cancer, coronaryheart disease, and type 2 diabetes.107–110

Additional avenues of biomarker iden-tification are actively being explored.There is growing interest in possiblebiologic measures that could be usedbefore (or immediately after) birth toassess risk for ASD. Such markers in-clude metabolites, amino acids, hor-mones, and immune factors, eitherindividually or in combination with thegoal of creating biomarker arrays toassess risk as well as severity, thusproviding information that could lead tospecific therapeutic interventions.111,112

Thus, future biomarker research shouldconsider how combinations of bio-markers could be used in prediction ofASD risk, and how incorporation ofbiomarker profiles together with be-havioral markers might improve onscreening methods based on themarkers alone. Although somemethodspresent logistical difficulties (eg, cost,invasiveness), others, such as EEGs, aremore readily available in pediatric set-tings (eg, auditory brainstem responseinnewborns),noninvasive,andrelativelyinexpensive. With further laboratoryand community-based research, suchmethods might ultimately exhibit thepotential to improve the sensitivity andspecificity of early detection, as well asenable detection earlier in development.

ACKNOWLEDGMENTSThe conference chairs and workinggroups acknowledge the preconferencecontribution of Grace Baranek, PhD,OTR/L, FAOTA, who was unable to attendthe conference. We also acknowledgethe efforts of Katherine F. Murray,BSN, RN, Massachusetts General Hospi-tal for Children, in coordinating the fo-rumandsubsequent conference reportprocess, and Sifor Ng in the conferencereport process.

The meeting and consensus reportwere sponsored by the Autism Forum.An important goal of the forum is toidentify early indicators of ASDs thatmay lead to effective health care ser-vices. Autism Forum programs are de-veloped under the guidance of itsparent organization, the Northwest Au-tism Foundation. For this project, theAutism Research Institute provided fi-nancial support.

REFERENCES

1. Zwaigenbaum L, Bauman ML, Stone W, et al.Early identification of autism spectrum dis-orders: Recommendations for practice andresearch. Pediatrics. 2015;136(suppl 1)

2. Johnson CP, Myers SM; American Acad-emy of Pediatrics Council on ChildrenWith Disabilities. Identification and evalu-ation of children with autism spectrum

disorders. Pediatrics. 2007;120(5):1183–1215

3. National Research Council. Committee onEducational Interventions for Children

SUPPLEMENT ARTICLE


with Autism, Division of Behavioral andSocial Sciences and Education. In: Lord C,McGee JP, eds. Educating Children WithAutism. Washington, DC: National Acad-emy Press; 2001

4. Council on Children With Disabilities;Section on Developmental Behavioral Pe-diatrics; Bright Futures Steering Commit-tee; Medical Home Initiatives for ChildrenWith Special Needs Project AdvisoryCommittee. Identifying infants and youngchildren with developmental disorders inthe medical home: an algorithm for de-velopmental surveillance and screening.Pediatrics. 2006;118(1):405–420

5. Dawson G, Rogers S, Munson J, et al.Randomized, controlled trial of an in-tervention for toddlers with autism: theEarly Start Denver Model. Pediatrics.2010;125(1). Available at: www.pediatrics.org/cgi/content/full/125/1/e17

6. Kasari C, Gulsrud AC, Wong C, Kwon S,Locke J. Randomized controlled caregivermediated joint engagement interventionfor toddlers with autism. J Autism DevDisord. 2010;40(9):1045–1056

7. Al-Qabandi M, Gorter JW, Rosenbaum P.Early autism detection: are we ready forroutine screening? Pediatrics. 2011;128(1). Available at: www.pediatrics.org/cgi/content/full/128/1/e211

8. Campos-Outcalt D. Should all children bescreened for autism spectrum disorders?No: screening is not ready for prime time.Am Fam Physician. 2011;84(4):377–378

9. Daniels AM, Halladay AK, Shih A, Elder LM,Dawson G. Approaches to enhancing theearly detection of autism spectrum dis-orders: a systematic review of the litera-ture. J Am Acad Child Adolesc Psychiatry.2014;53(2):141–152

10. Roux AM, Herrera P, Wold CM, Dunkle MC,Glascoe FP, Shattuck PT. Developmentaland autism screening through 2-1-1:reaching underserved families. Am J PrevMed. 2012;43(6 suppl 5):S457–S463

11. Dosreis S, Weiner CL, Johnson L, News-chaffer CJ. Autism spectrum disorderscreening and management practicesamong general pediatric providers. J DevBehav Pediatr. 2006;27(suppl 2):S88–S94

12. King TM, Tandon SD, Macias MM, et al.Implementing developmental screeningand referrals: lessons learned from a na-tional project. Pediatrics. 2010;125(2):350–360

13. Pinto-Martin JA, Dunkle M, Earls M,Fliedner D, Landes C. Developmentalstages of developmental screening: stepsto implementation of a successful pro-

gram. Am J Public Health. 2005;95(11):1928–1932

14. Koegel LK, Koegel RL, Nefdt N, Fredeen R,Klein EF, Bruinsma YE. First STEP: a modelfor the early identification of childrenwith autism spectrum disorders. J PositBehav Interv. 2005;7(4):247–252

15. Warren Z, Stone W, Humberd Q. A trainingmodel for the diagnosis of autism incommunity pediatric practice. J DevBehav Pediatr. 2009;30(5):442–446

16. Webb PL. Screening for autism spectrumdisorders during well-child visits in a pri-mary care setting. J Nurse Pract. 2011;7(3):229–235

17. Pierce K, Carter C, Weinfeld M, et al.Detecting, studying, and treating autismearly: the one-year well-baby check-upapproach. J Pediatr. 2011;159(3):458–465.e1–e6

18. Lipkin PH, Hyman SL. Should all childrenbe screened for autism spectrum dis-orders? Yes: merging science, policy, andpractice. Am Fam Physician. 2011;84(4):361–367

19. Dumont-Mathieu T, Fein D. Screening forautism in young children: the ModifiedChecklist for Autism in Toddlers (M-CHAT)and other measures. Ment Retard DevDisabil Res Rev. 2005;11(3):253–262

20. Reznick JS, Baranek GT, Reavis S, WatsonLR, Crais ER. A parent-report instrumentfor identifying one-year-olds at risk for aneventual diagnosis of autism: the firstyear inventory. J Autism Dev Disord. 2007;37(9):1691–1710

21. Turner-Brown LM, Baranek GT, Reznick JS,Watson LR, Crais ER. The First Year In-ventory: a longitudinal follow-up of 12-month-old to 3-year-old children. Autism.2013;17(5):527–540

22. Dietz C, Swinkels S, van Daalen E, vanEngeland H, Buitelaar JK. Screening forautistic spectrum disorder in childrenaged 14-15 months. II: Population screen-ing with the Early Screening of AutisticTraits Questionnaire (ESAT). Design andgeneral findings. J Autism Dev Disord.2006;36(6):713–722

23. Swinkels SH, Dietz C, van Daalen E, Ker-khof IH, van Engeland H, Buitelaar JK.Screening for autistic spectrum in chil-dren aged 14 to 15 months. I: The devel-opment of the Early Screening of AutisticTraits Questionnaire (ESAT). J Autism DevDisord. 2006;36(6):723–732

24. Robins DL, Fein D, Barton ML, Green JA.The Modified Checklist for Autism in Tod-dlers: an initial study investigating theearly detection of autism and pervasive

developmental disorders. J Autism DevDisord. 2001;31(2):131–144

25. Robins DL. Screening for autism spectrumdisorders in primary care settings. Au-tism. 2008;12(5):537–556

26. Kleinman JM, Robins DL, Ventola PE, et al.The Modified Checklist for Autism in Tod-dlers: a follow-up study investigating theearly detection of autism spectrum dis-orders. J Autism Dev Disord. 2008;38(5):827–839

27. Pandey J, Verbalis A, Robins DL, et al.Screening for autism in older and youn-ger toddlers with the Modified Checklistfor Autism in Toddlers. Autism. 2008;12(5):513–535

28. Inada N, Koyama T, Inokuchi E, Kuroda M,Kamio Y. Reliability and validity of theJapanese version of the Modified Check-list for Autism in Toddlers (M-CHAT). ResAutism Spectr Disord. 2011;5(1):330–336

29. Canal-Bedia R, García-Primo P, Martín-Cil-leros MV, et al. Modified checklist for au-tism in toddlers: cross-cultural adaptationand validation in Spain. J Autism DevDisord. 2011;41(10):1342–1351

30. Pinto-Martin JA, Young LM, Mandell DS,Poghosyan L, Giarelli E, Levy SE. Screeningstrategies for autism spectrum disordersin pediatric primary care. J Dev BehavPediatr. 2008;29(5):345–350

31. Chlebowski C, Robins DL, Barton ML, FeinD. Large-scale use of the modified check-list for autism in low-risk toddlers. Pedi-atrics. 2013;131(4). Available at: www.pediatrics.org/cgi/content/full/131/4/e1121

32. Robins DL, Casagrande K, Barton M, ChenCM, Dumont-Mathieu T, Fein D. Validationof the Modified Checklist for Autism inToddlers, Revised With Follow-Up (M-CHAT-R/F). Pediatrics. 2014;133(1):37–45

33. Allison C, Baron-Cohen S, Wheelwright S,et al.The Q-CHAT (Quantitative Checklist forAutism in Toddlers): a normally distrib-uted quantitative measure of autistictraits at 18-24 months of age: preliminaryreport. J Autism Dev Disord. 2008;38(8):1414–1425

34. Allison C, Auyeung B, Baron-Cohen S. To-ward brief “Red Flags” for autismscreening: the Short Autism SpectrumQuotient and the Short QuantitativeChecklist for Autism in toddlers in 1,000cases and 3,000 controls [published cor-rection appears in J Am Acad Child Ado-lesc Psychiatry. 2012;51(3):338]. J AmAcad Child Adolesc Psychiatry. 2012;51(2):202–212.e7

35. Gray KM, Tonge BJ, Sweeney DJ, Einfeld SL.Screening for autism in young children


with developmental delay: an evaluationof the developmental behaviour checklist:early screen. J Autism Dev Disord. 2008;38(6):1003–1010

36. Eaves RC, Williams TO Jr. Exploratory andconfirmatory factor analyses of the per-vasive developmental disorders ratingscale for young children with autisticdisorder. J Genet Psychol. 2006;167(1):65–92

37. Eaves RC, Woods-Groves S, Williams TO Jr,Fall AM. Reliability and validity of thePervasive Developmental Disorders Rat-ing Scale and the Gilliam Autism RatingScale. Educ Train Dev Disabil. 2006;41(3):300–309

38. Stone WL, Coonrod EE, Ousley OY. Briefreport: Screening Tool for Autism in Two-year-olds (STAT): development and pre-liminary data. J Autism Dev Disord. 2000;30(6):607–612

39. Stone WL, Coonrod EE, Turner LM, PozdolSL. Psychometric properties of the STATfor early autism screening. J Autism DevDisord. 2004;34(6):691–701

40. Stone WL, McMahon CR, Henderson LM.Use of the Screening Tool for Autism inTwo-Year-Olds (STAT) for children under 24months: an exploratory study. Autism.2008;12(5):557–573

41. Matson JL, Wilkins J, Sevin JA, Knight C,Boisjoli JA, Sharp B. Reliability and itemcontent of the Baby and Infant Screen forChildren with Autism Traits (BISCUIT):parts 1–3. Res Autism Spectr Disord.2009;3(2):336–344

42. Matson JL, Wilkins J, Sharp B, KnightC, Sevin JA, Boisjoli JA. Sensitivity andspecificity of the Baby and InfantScreen for Children with Autism Traits(BISCUIT): validity and cutoff scoresfor autism and PDD-NOS in toddlers.Res Autism Spectr Disord. 2009;3(4):924–930

43. Wetherby AM, Woods J, Allen L, Cleary J,Dickinson H, Lord C. Early indicators ofautism spectrum disorders in the secondyear of life. J Autism Dev Disord. 2004;34(5):473–493

44. Wetherby AM, Brosnan-Maddox S, Peace V,Newton L. Validation of the Infant-ToddlerChecklist as a broadband screener forautism spectrum disorders from 9 to 24months of age. Autism. 2008;12(5):487–511

45. Oosterling IJ, Swinkels SH, van der GaagRJ, Visser JC, Dietz C, Buitelaar JK. Com-parative analysis of three screeninginstruments for autism spectrum disor-der in toddlers at high risk. J Autism DevDisord. 2009;39(6):897–909

46. Baron-Cohen S, Allen J, Gillberg C. Canautism be detected at 18 months? Theneedle, the haystack, and the CHAT. Br JPsychiatry. 1992;161:839–843

47. Baron-Cohen S, Cox A, Baird G, et al. Psy-chological markers in the detection ofautism in infancy in a large population. BrJ Psychiatry. 1996;168(2):158–163

48. Watson LR, Baranek GT, Crais ER, StevenReznick J, Dykstra J, Perryman T. The firstyear inventory: retrospective parent re-sponses to a questionnaire designed toidentify one-year-olds at risk for autism. JAutism Dev Disord. 2007;37(1):49–61

49. Baird G, Charman T, Baron-Cohen S, et al.A screening instrument for autism at 18months of age: a 6-year follow-up study. JAm Acad Child Adolesc Psychiatry. 2000;39(6):694–702

50. Albores-Gallo L, Roldán-Ceballos O,Villarreal-Valdes G, et al. M-CHAT Mexicanversion validity and reliability and somecultural considerations. ISRN Neurol.2012;2012:408694

51. Seif Eldin A, Habib D, Noufal A, et al. Use ofM-CHAT for a multinational screening ofyoung children with autism in the Arabcountries. Int Rev Psychiatry. 2008;20(3):281–289

52. Kobak KA, Stone WL, Ousley OY, Swanson A.Web-based training in early autismscreening: results from a pilot study. Tel-emed J E Health. 2011;17(8):640–644

53. Wetherby A, Prizant B. Communicationand Symbolic Behavior Scales De-velopmental Profile—First Normed Edi-tion. Baltimore, MD: Paul H. BrookesPublishing Co; 2002

54. Wiggins LD, Piazza V, Robins DL. Compar-ison of a broad-based screen versusdisorder-specific screen in detectingyoung children with an autism spectrumdisorder. Autism. 2014;18(2):76–84

55. Glascoe FP. Parents’ Evaluation of De-velopmental Status. Nashville, TN: Ells-worth & Vandermeer Press, Ltd; 1997

56. Bricker D, Squires J. Ages & StagesQuestionnaires: A Parent-Completed, Child-Monitoring System. 2nd ed. Baltimore, MD:Paul H. Brookes Publishing Co; 1999

57. Constantino JN, Zhang Y, Frazier T,Abbacchi AM, Law P. Sibling recurrenceand the genetic epidemiology of autism.Am J Psychiatry. 2010;167(11):1349–1356

58. Ozonoff S, Young GS, Carter A, et al. Re-currence risk for autism spectrum dis-orders: a Baby Siblings ResearchConsortium study. Pediatrics. 2011;128(3).Available at: www.pediatrics.org/cgi/con-tent/full/128/3/e488

59. Grønborg TK, Schendel DE, Parner ET. Re-currence of autism spectrum disorders infull- and half-siblings and trends overtime: a population-based cohort study.JAMA Pediatr. 2013;167(10):947–953

60. Ritvo ER, Jorde LB, Mason-Brothers A,et al. The UCLA-University of Utah epide-miologic survey of autism: recurrencerisk estimates and genetic counseling. AmJ Psychiatry. 1989;146(8):1032–1036

61. Georgiades S, Szatmari P, Zwaigenbaum L,et al. A prospective study of autistic-liketraits in unaffected siblings of probandswith autism spectrum disorder. JAMAPsychiatry. 2013;70(1):42–48

62. Landa RJ, Holman KC, Garrett-Mayer E.Social and communication developmentin toddlers with early and later diagnosisof autism spectrum disorders. Arch GenPsychiatry. 2007;64(7):853–864

63. Messinger D, Young GS, Ozonoff S, et al.Beyond autism: a baby siblings researchconsortium study of high-risk children atthree years of age. J Am Acad Child Ado-lesc Psychiatry. 2013;52(3):300–308.e1

64. Stone WL, McMahon CR, Yoder PJ, WaldenTA. Early social-communicative and cog-nitive development of younger siblings ofchildren with autism spectrum disorders.Arch Pediatr Adolesc Med. 2007;161(4):384–390

65. van Daalen E, Kemner C, Dietz C, SwinkelsSH, Buitelaar JK, van Engeland H. Inter-rater reliability and stability of di-agnoses of autism spectrum disorder inchildren identified through screening ata very young age. Eur Child Adolesc Psy-chiatry. 2009;18(11):663–674

66. Kleinman JM, Ventola PE, Pandey J, et al.Diagnostic stability in very young childrenwith autism spectrum disorders. J AutismDev Disord. 2008;38(4):606–615

67. Chawarska K, Klin A, Paul R, Volkmar F.Autism spectrum disorder in the secondyear: stability and change in syndromeexpression. J Child Psychol Psychiatry.2007;48(2):128–138

68. Gillberg C, Ehlers S, Schaumann H, et al.Autism under age 3 years: a clinical studyof 28 cases referred for autistic symp-toms in infancy. J Child Psychol Psychia-try. 1990;31(6):921–934

69. Charman T, Taylor E, Drew A, Cockerill H,Brown JA, Baird G. Outcome at 7 years ofchildren diagnosed with autism at age 2:predictive validity of assessments con-ducted at 2 and 3 years of age and patternof symptom change over time. J ChildPsychol Psychiatry. 2005;46(5):500–513

70. Cox A, Klein K, Charman T, et al. Autismspectrum disorders at 20 and 42 months

SUPPLEMENT ARTICLE


of age: stability of clinical and ADI-R di-agnosis. J Child Psychol Psychiatry. 1999;40(5):719–732

71. Lord C. Follow-up of two-year-olds re-ferred for possible autism. J Child PsycholPsychiatry. 1995;36(8):1365–1382

72. Lord C, Risi S, DiLavore PS, Shulman C,Thurm A, Pickles A. Autism from 2 to 9years of age. Arch Gen Psychiatry. 2006;63(6):694–701

73. Stone WL, Lee EB, Ashford L, et al. Canautism be diagnosed accurately in chil-dren under 3 years? J Child Psychol Psy-chiatry. 1999;40(2):219–226

74. Turner LM, Stone WL. Variability in out-come for children with an ASD diagnosisat age 2. J Child Psychol Psychiatry. 2007;48(8):793–802

75. Gillis JM. Screening practices of familyphysicians and pediatricians in 2 south-ern states. Infants Young Child. 2009;22(4):321–331

76. Miller JS, Gabrielsen T, Villalobos M, et al.The each child study: systematic screen-ing for autism spectrum disorders ina pediatric setting. Pediatrics. 2011;127(5):866–871

77. McLennan JD, Huculak S, Sheehan D. Briefreport: pilot investigation of service re-ceipt by young children with autisticspectrum disorders. J Autism Dev Disord.2008;38(6):1192–1196

78. Kennedy T, Regehr G, Rosenfield J, RobertsSW, Lingard L. Exploring the gap betweenknowledge and behavior: a qualitativestudy of clinician action following an ed-ucational intervention. Acad Med. 2004;79(5):386–393

79. Daniels AM, Mandell DS. Explaining dif-ferences in age at autism spectrum dis-order diagnosis: a critical review. Autism.2014;18(5):583–597

80. Magaña S, Lopez K, Aguinaga A, Morton H.Access to diagnosis and treatment ser-vices among Latino children with autismspectrum disorders. Intellect Dev Disabil.2013;51(3):141–153

81. Zuckerman KE, Mattox K, Donelan K, Bat-bayar O, Baghaee A, Bethell C. Pediatricianidentification of Latino children at risk forautism spectrum disorder. Pediatrics.2013;132(3):445–453

82. Herlihy LE, Brooks B, Dumont-Mathieu T,et al. Standardized screening facilitatestimely diagnosis of autism spectrum dis-orders in a diverse sample of low-risktoddlers. J Dev Behav Pediatr. 2014;35(2):85–92

83. Filipek PA, Accardo PJ, Ashwal S, et al.Practice parameter: screening and di-agnosis of autism: report of the quality

standards subcommittee of the AmericanAcademy of Neurology and the ChildNeurology Society. Neurology. 2000;55(4):468–479

84. Oosterling I, Visser J, Swinkels S, et al.Randomized controlled trial of the focusparent training for toddlers with autism:1-year outcome. J Autism Dev Disord.2010;40(12):1447–1458

85. Prorok PC, Marcus PM. Cancer screeningtrials: nuts and bolts. Semin Oncol. 2010;37(3):216–223

86. Colditz GA, Taylor PR. Prevention trials:their place in how we understand thevalue of prevention strategies. Annu RevPublic Health. 2010;31:105–120

87. Kogan MD, Strickland BB, Blumberg SJ,Singh GK, Perrin JM, van Dyck PC. A na-tional profile of the health care experi-ences and family impact of autismspectrum disorder among children in theUnited States, 2005-2006. Pediatrics. 2008;122(6). Available at: www.pediatrics.org/cgi/content/full/122/6/e1149

88. Leslie DL, Martin A. Health care expendi-tures associated with autism spectrumdisorders. Arch Pediatr Adolesc Med.2007;161(4):350–355

89. Liptak GS, Stuart T, Auinger P. Health careutilization and expenditures for childrenwith autism: data from U.S. nationalsamples. J Autism Dev Disord. 2006;36(7):871–879

90. Harrington JW, Bai R, Perkins AM.Screening children for autism in an urbanclinic using an electronic M-CHAT. ClinPediatr (Phila). 2013;52(1):35–41

91. Heidgerken AD, Geffken G, Modi A, FrakeyL. A survey of autism knowledge ina health care setting. J Autism Dev Disord.2005;35(3):323–330

92. Hebert EB, Koulouglioti C. Parental beliefsabout cause and course of their child’sautism and outcomes of their beliefs:a review of the literature. Issues ComprPediatr Nurs. 2010;33(3):149–163

93. Bernier R, Mao A, Yen J. Psychopathology,families, and culture: autism. Child Ado-lesc Psychiatr Clin N Am. 2010;19(4):855–867

94. Cadman D, Chambers LW, Walter SD,Ferguson R, Johnston N, McNamee J.Evaluation of public health preschoolchild developmental screening: the pro-cess and outcomes of a communityprogram. Am J Public Health. 1987;77(1):45–51

95. Harrington JW, Patrick PA, Edwards KS,Brand DA. Parental beliefs about autism:implications for the treating physician.Autism. 2006;10(5):452–462

96. Courchesne E, Pierce K, Schumann CM,et al. Mapping early brain development inautism. Neuron. 2007;56(2):399–413

97. Nickl-Jockschat T, Habel U, Michel TM,et al. Brain structure anomalies in autismspectrum disorder—a meta-analysis ofVBM studies using anatomic likelihoodestimation. Hum Brain Mapp. 2012;33(6):1470–1489

98. Dinstein I, Pierce K, Eyler L, et al. Dis-rupted neural synchronization in toddlerswith autism. Neuron. 2011;70(6):1218–1225

99. Bosl W, Tierney A, Tager-Flusberg H, Nel-son C. EEG complexity as a biomarker forautism spectrum disorder risk. BMC Med.2011;9:18

100. Elsabbagh M, Mercure E, Hudry K, et al;BASIS Team. Infant neural sensitivity todynamic eye gaze is associated with lateremerging autism. Curr Biol. 2012;22(4):338–342

101. Wolff JJ, Gu H, Gerig G, et al; IBIS Network.Differences in white matter fiber tractdevelopment present from 6 to 24 monthsin infants with autism. Am J Psychiatry.2012;169(6):589–600

102. Sebat J, Lakshmi B, Malhotra D, et al.Strong association of de novo copy num-ber mutations with autism. Science. 2007;316(5823):445–449

103. Yu TW, Chahrour MH, Coulter ME, et al.Using whole-exome sequencing to identifyinherited causes of autism. Neuron. 2013;77(2):259–273

104. Jiang YH, Yuen RK, Jin X, et al. Detection ofclinically relevant genetic variants in au-tism spectrum disorder by whole-genomesequencing. Am J Hum Genet. 2013;93(2):249–263

105. Carayol J, Schellenberg GD, Tores F,Hager J, Ziegler A, Dawson G. Assess-ing the impact of a combined analysisof four common low-risk genetic var-iants on autism risk. Mol Autism. 2010;1(1):4

106. Pierce K, Glatt SJ, Liptak GS, McIntyre LL.The power and promise of identifyingautism early: insights from the search forclinical and biological markers. Ann ClinPsychiatry. 2009;21(3):132–147

107. Zethelius B, Berglund L, Sundström J,et al. Use of multiple biomarkers to im-prove the prediction of death from car-diovascular causes. N Engl J Med. 2008;358(20):2107–2116

108. Morrison AC, Bare LA, Chambless LE, et al.Prediction of coronary heart disease riskusing a genetic risk score: the Athero-sclerosis Risk in Communities Study. Am JEpidemiol. 2007;166(1):28–35


109. Weedon MN, McCarthy MI, Hitman G, et al.Combining information from common type 2diabetes risk polymorphisms improves dis-ease prediction. PLoS Med. 2006;3(10):e374

110. Yang Q, Khoury MJ, Botto L, Friedman JM,Flanders WD. Improving the prediction of

complex diseases by testing for multipledisease-susceptibility genes. Am J HumGenet. 2003;72(3):636–649

111. Ratajczak HV. Theoretical aspects of au-tism: biomarkers—a review. J Immuno-toxicol. 2011;8(1):80–94

112. James SJ, Cutler P, Melnyk S, et al.Metabolic biomarkers of increasedoxidative stress and impaired meth-ylation capacity in children with au-tism. Am J Clin Nutr. 2004;80(6):1611–1617

(Continued from first page)

ABBREVIATIONSAAP—American Academy of PediatricsASD—autism spectrum disorderCHAT—Checklist for Autism in ToddlersCSBS DP—Communication and Symbolic Behavior Scales Developmental ProfileESAT—Early Screening of Autistic Traits questionnaireFYI—First Year InventoryM-CHAT—Modified Checklist for Autism in ToddlersNPV—negative predictive valuePEDS—Parents’ Evaluation of Developmental StatusPPV—positive predictive valueQ-CHAT—Quantitative Checklist for Autism in ToddlersSTAT—Screening Tool for Autism in Two-Year-Olds

Drs Zwaigenbaum and Bauman initiated a literature review, co-chaired the meeting that generated the consensus recommendations outlined in this article, anddrafted the initial manuscript; Drs Fein and Pierce co-chaired the working group that conducted the detailed literature review, generated initial recommendationsthat were discussed at the consensus meeting, and provided critical input to subsequent drafts of the manuscript; Drs Buie, Davis, Newschaffer, Robins, andWetherby were members of the working group that reviewed selected publications, contributed to initial recommendations that were reviewed at the consensusmeeting, and critically reviewed the manuscript; and Drs Choureiri, Kasari, Stone, Yirmiya, Estes, Hansen, McPartland, Natowicz, Carter, Granpeesheh, Mailloux,Smith Roley, and Wagner contributed to the consensus meeting that formed the basis for the manuscript and critically reviewed the manuscript. All authorsapproved the final manuscript as submitted.

www.pediatrics.org/cgi/doi/10.1542/peds.2014-3667D

doi:10.1542/peds.2014-3667D

Accepted for publication Aug 3, 2015

Address correspondence to Lonnie Zwaigenbaum, MD, Autism Research Center, Glenrose Rehabilitation Hospital, Room E209, 10230 111 Ave, Edmonton, AB, CanadaT5G 0B7. E-mail: [email protected]

PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275).

Copyright © 2015 by the American Academy of Pediatrics

FINANCIAL DISCLOSURE: Dr Zwaigenbaum was the site Principal Investigator of a study sponsored by SynapDx (he received operating funds but no honoraria).Drs Fein and Robins are co-owners of M-CHAT, LLC, which licenses use of the Modified Checklist for Autism in Toddlers in electronic products. Dr Stone is the authorof the Screening Tool for Autism in Two-Year-Olds and receives a share of royalties from sales of this instrument. The authors received an honorarium as well astravel expenses from Autism Forum for contributing to the expert panels.

FUNDING: Sponsored by the Autism Forum under the guidance of the Northwest Autism Foundation and with the support of the Autism Research Institute.

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

SUPPLEMENT ARTICLE


mailto:[email protected]

DOI: 10.1542/peds.2014-3667D2015;136;S41Pediatrics

Granpeesheh, Zoe Mailloux, Susanne Smith Roley and Sheldon WagnerHansen, James C. McPartland, Marvin R. Natowicz, Alice Carter, Doreen

Choueiri, Connie Kasari, Wendy L. Stone, Nurit Yirmiya, Annette Estes, Robin L.Buie, Patricia A. Davis, Craig Newschaffer, Diana L. Robins, Amy Wetherby, Roula Lonnie Zwaigenbaum, Margaret L. Bauman, Deborah Fein, Karen Pierce, Timothy

and ResearchEarly Screening of Autism Spectrum Disorder: Recommendations for Practice

ServicesUpdated Information &

http://pediatrics.aappublications.org/content/136/Supplement_1/S41including high resolution figures, can be found at:

References

BIBLhttp://pediatrics.aappublications.org/content/136/Supplement_1/S41#This article cites 105 articles, 16 of which you can access for free at:

Permissions & Licensing

http://www.aappublications.org/site/misc/Permissions.xhtmlin its entirety can be found online at: Information about reproducing this article in parts (figures, tables) or

Reprintshttp://www.aappublications.org/site/misc/reprints.xhtmlInformation about ordering reprints can be found online:


http://http://pediatrics.aappublications.org/content/136/Supplement_1/S41

http://pediatrics.aappublications.org/content/136/Supplement_1/S41#BIBL

http://pediatrics.aappublications.org/content/136/Supplement_1/S41#BIBL

http://www.aappublications.org/site/misc/Permissions.xhtml

http://www.aappublications.org/site/misc/reprints.xhtml

DOI: 10.1542/peds.2014-3667D2015;136;S41Pediatrics

Granpeesheh, Zoe Mailloux, Susanne Smith Roley and Sheldon WagnerHansen, James C. McPartland, Marvin R. Natowicz, Alice Carter, Doreen

Choueiri, Connie Kasari, Wendy L. Stone, Nurit Yirmiya, Annette Estes, Robin L.Buie, Patricia A. Davis, Craig Newschaffer, Diana L. Robins, Amy Wetherby, Roula Lonnie Zwaigenbaum, Margaret L. Bauman, Deborah Fein, Karen Pierce, Timothy

and ResearchEarly Screening of Autism Spectrum Disorder: Recommendations for Practice

http://pediatrics.aappublications.org/content/136/Supplement_1/S41located on the World Wide Web at:

The online version of this article, along with updated information and services, is

ISSN: 1073-0397. 60007. Copyright © 2015 by the American Academy of Pediatrics. All rights reserved. Print the American Academy of Pediatrics, 141 Northwest Point Boulevard, Elk Grove Village, Illinois,has been published continuously since 1948. Pediatrics is owned, published, and trademarked by Pediatrics is the official journal of the American Academy of Pediatrics. A monthly publication, it


http://pediatrics.aappublications.org/content/136/Supplement_1/S41

Early Screening of Autism Spectrum Disorder ... · mote earlier diagnosis. The current review indicates that screening in children aged 18 to 24 months can assist in early detection,

Documents